Specific satellite image spectrum can well perform automatically classification for monitoring agricultural land cover, but the obstacle for urban area land use classification application is still under research stage. The fundamental barrier is due to the fact that land use for urban area is composed by several groups of complex land cover types which reflect differently from any single land cover. Traditional pixel based classification assign each pixel a certain land cover type from its spectrum reflectance value and through statistical analysis. The process ignores neighboring relationship between each pixel spectrum reflectance which cause the main reason for poor land use classification for urban area. This study applied Spatial Post Classification Algorithm on high resolution satellite image for urban land use classification. The classification can be disaggregated into two stages, the first stage is standard per pixel based classification. At this stage, the land cover was identified as seven types (large construction, small construction, forest, crop, grass, bare, and water) by Gauss Maximum Likelihood method. The results showed an overall accuracy to 92.7% with 0.91 Kappa value. The second stage applied seven different window matrix sizes (3*3, 5*5, 7*7, 9*9, 11*11, 21*21, and 31*31) as post classification from first stage outcomes. The land use can then be identified as eight types (high density housing, middle density housing, industrial, farm, pasture, forest, open space, and water). The result showed an optimum window matrix size was 11*11 for complicated and fractal urban area land use type classification. This window size can well describe spatial changes of land cover through entire study area, and can minimize smooth effect from large window matrix size. The final land use classification can reach overall accuracy to 89.6% with 0.86 Kappa value.

Changes in normalized difference vegetation index (NDVI), which calculated from ground-based canopy hyperspectral reflectance data, and leaf area index (LAI), which measured at the time of spectral measurements, were monitored during rice (Oryza sativa L. cv. TNG 67) growth and yields were harvested at maturity so as to determine the NDVI-LAI relationships and the optimal timing to use these two parameters for yield prediction. Field experiments were conducted at the experimental farm of Taiwan Agricultural Research Institute (Wufeng, Taiwan) in the first and the second cropping seasons of 2001-2002. Different levels of nitrogen fertilizer, from 0 to 180 kg ha-1 with 30 kg ha-1 intervals, were applied to produce various scales of grain yield and values of LAI and NDVI. Results showed that there were differential weather conditions between two cropping seasons, and plants grown in first crops had a better growth and yield. Both LAI and NDVI were found curvilinearly distributed during the growing periods, with the maximum values occurred before heading in First Crops and near or after heading in Second Crops. These two parameters can be mutually estimated through an exponential function, which linked spectral remote sensing data with plant growth information. Correlation between yield and LAI was best fitted to a nonlinear function, from about 50 days after transplanting (DAT) with the determining factor (R2) higher than 0.73 (P<0.01). By further analyzing the relations of the cumulative and the mean values of LAI to yield, it indicated that both the accumulated and the mean values of LAI from 15 days before heading (DBH) to 15 days after heading (DAH) can well predict rice yield for the first cropping seasons, while values calculated from 15 BH to 10 DAH can predict yield for the second cropping seasons.

For this study we used a CCD video camera to capture the pavement image information via the computer. During investigation processing, the CCD video camera captured 10~30 images per second. If the vehicle velocity is too fast, the collected images will be duplicated and if the velocity is too slow there will be a gapped between images. Therefore, in order to control the efficiency of the image grabber we should add accessory tools such as the Differential Global Positioning System (DGPS) and odometer. Furthermore, Kalman Filtering can also solve these problems. After the CCD video camera captured the pavement images, we used the Least-Squares method to eliminate images of gradation which have non-uniform surfaces due to the illumination at night. The Fuzzy Entropy method calculates images of threshold segments and creates binary images. Finally, the Object Labeling algorithm finds objects that are cracks or noises from the binary image based on volume pixels of the object. We used these algorithms and tested them, also providing some discussion and suggestions.

Digital surface model generation from IKONOS stereo imagery is a new challenge in photogrammetric community, especially when the satellite company does not provide the raw data as well as their ancillary data. The IKONOS stereo images are provided in a quasi-epipolar geometry reference where only the terrain parallax in the scanner direction remains. A rigorous solution for orbit modeling is not practical. In this paper we utilized an estimated relief displacement azimuth and the nominal collection elevation data included in the metadata file to correct the relief displacement of GCPs, together with a linear transformation for geometric modeling of IKONOS imagery. Conjugate point correspondence problem is solved by a multi-resolution template-based image matching technique. Space intersection is performed by the trigonometric intersection assuming a parallel projection of IKONOS imagery due to its small FOV and frame size. In the experiment, less than 2-meters of RMSE is achieved denoting the potential positioning accuracy of the IKONOS stereo imagery.

It is potentially helpful to improve the weather forecasting through assimilating the atmospheric wet refractivity measured with Global Positioning System (GPS) into numerical weather prediction models. In this paper, a tomographic method is developed to reconstruct the 3D wet refractivity structures of the troposphere using ground-based GPS measured slant delay based on an observing system simulation experiment (OSSE)-type approach. The slant delay is derived from a reference atmospheric temperature, pressure, and water vapor profiles (or “ground truth”) measured by a multichannel microwave radiometer, which subsequently serve as reference for comparison with the reconstructed 3D wet delay distribution. The reconstruction is possible because a ground network of many GPS receivers that record integrated slant water along the ray paths of the electromagnetic waves provide the required information. In the simulation, a network with 5 by 5 GPS receivers is uniformly distributed. Distance between two conses (5 by 5 by 11 unknown points). The GPS signals that pass through the 160 cubes intersect all over during a certain time interval of interest, and permit the reconstruction. The relative errors of the reconstructed wet refractivity can be as low as 5 to 10% below 3-5 km altitude compared with the ground truth. They can be further decreased when surface meteorological measurements and an assumption of null refractivity at a reasonable height are applied as constraints. In addition, positional dilution of precision (PDOP) factors is computed to assist in the analysis of the results.

This paper presents accuracy assessment of InSAR derived DEM acquired by Intermap’s IFSAR system. The flight area was located southwestern Taiwan covering counties of Chiayi, Tainan, Kaohsiung and Pingtung with about 3676 squared km mapping size. The output data included ortho-rectified radar image at 1.25 m resolution and digital elevation model at 5 meter posting. We compared the DEM with 292 ground control points provided by Satellite Survey Center, Department of Land Administration, MOI. It was found that the height difference standard deviation was around 1.5 m and the bias was 0.6 m. In this experiment campaign, the accuracy obtained from InSAR derived is highly acceptable for many applications. The error sources were attributed to the difference between the use of GPS point height and InSAR area (averaged over 5 squared meters) height. The relation of height difference to aspect and slope illustrated that there exist weak dependence of terrain effects. For purpose of demonstration, we overlaid the flooded area derived from RADARSAT-1 with DEM at fine and coarse resolution to illustrate the importance of high resolution DEM data in assessing the flood risk.

Satellite images are advantageous to large-area and repetitive observation, and therefore, are gradually adopted in the tasks of resource monitoring. Their analysis and presentation for the present are important to resource monitoring, but the most widely used SPOT satellite images can only show the image of false color, not the true color that is familiar to human eyes. In an attempt to enhance the visual effect of SPOT image, this study develops a neural network algorithm to transform SPOT false color into simulated true color. The method has been tested using Landsat TM and SPOT images. The qualitative and quantitative comparisons indicate that the striking similarity can be found between the true and simulated true images in terms of the visual looks and the statistical analysis.

Wavelet based image compression techniques can provide high compression ratio with high fidelity. Besides the JPEG2000 format, there are also several commercial software for image compression which are based on the wavelet theory. This study investigates the performance of Jasper, the reference software of JPEG2000, and two commercial software, LuraWave and ECW, with a scanned grey-scale aerial photograph and a color image. The objective of this study is not to determine which software or format is better than the others, but to visualize the difference between different formats which are all based on the wavelet theory. With the study on limited samples, this study hopes to establish an evaluation scheme for the image compression utilities.